Table 1 Review on methods and quantitative results for the classification of COVID-19 CT-Scan Images.

Zhao et al., 2021¹⁰

To make use of CNNs in combination with transfer learning techniques

This study uses the COVIDx CT-2 dataset

The pretrained ImageNet21k model is employed. The tSNE nonlinear dimensionality reduction approach

Demonstrates an increase in the classification accuracy of CT-Scan images taken from out-of-field datasets

Silva et al., 2020¹¹

To improve the accuracy of Effiecint CovidNet model performance

Multiple datasets were retrieved from data repositories and journals

Effiecint CovidNet integrating with voting-based technique

This study shows an improved accuracy of 87.68%

Li et al., 2021¹²

To perform multi-classification prediction challenges

CT scans of 1417 patients

Present a technique for cascading classifiers that combine Stacked ensemble learning with VGG16

Accuracy:93.5 sensitivity:94.2, specificity: 93.9 F1-score: 91.7

Halder et al., 2021¹³

To determine the most appropriate model for COVID CT Scan classification using transfer learning techniques

CT scans of patients were taken from hospitals in São Paulo, Brazil through Kaggle

Transfer learning models–VGG16, DenseNet201, ResNet50V2, and MobileNet

Accuracies of DenseNet201 ResNet50V2, MobileNet, and VGG16 are 97%, 96%, 95% and 94% respectively

Wang et al., 2021¹⁴

Classifying CT Scan images by extracting COVID-19-specific graphical features

This study examines 1065 CT scans of COVID-19 cases with pathogen confirmation from three different hospitals

Multiple pre-processing techniques followed by M-Inception transfer learning model is used

Accuracy was 82.5 percent, sensitivity was 0.75, specificity was 0.86, PPV was 0.69, NPV was 0.89, and kappa was 0.59

Shah et al., 2021¹⁵

To find out the best suitable deep learning model for the COVID CT Scan image classification

The COVID-CT-Dataset contains a total of 738 people’s CT reports

This study has proposed a new model named CTnet-10 and compared the results with six different transfer learning models

Accuracy of CTnet-10 model was 82.1%. But, VGG-19 model surpassed with an accuracy of 94.52%

Mukherjee et al., 2021¹⁶

To design a NN that best suitable for both CT and CXR types of COVID medical images

The mix of 672 CT Scan and CXR images

A customized neural net CNN along with the DNN is used to train/test both the types of images

The obtained accuracy is 96.28%, AUC is 0.9808

Pham, 2020¹⁷

To investigate the 16 pretrained CNNs for COVID-19 classification

349 CT images from COVID-19 patients and 397 CT images from non-COVID subjects

Sixteen pretrained CNNs were tested using raw data and augmented data separately

DenseNet-201 model has highest accuracy and AUC